Flood depth extraction method and system based on remote sensing image, and electronic equipment

Abstract

The invention discloses a flood depth extraction method and system based on a remote sensing image and electronic equipment, and the method comprises the steps: carrying out the water extraction based on SAR image data of a target region and a trained water extraction model, so as to obtain a water extraction result; obtaining original river channel data, and superposing the original river channel data and the water body extraction result to obtain a flooding area; obtaining original DEM data of the target area, establishing a hydrological relationship based on the original DEM data, and obtaining a nearest neighbor drainage network relative height map of the target area; and determining the flood depth of the flooding area based on the relative height map of the nearest neighbor drainage network, and generating a water body depth map of the flooding area. According to the method, the flood depth can be efficiently and accurately determined, the flooding area water body depth map is generated, and the method has multiple advantages in the aspects of calculation efficiency, extraction precision, application range, data dependence, operation difficulty and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and more specifically, to a method, system and electronic equipment for extracting flood depth based on remote sensing images. Background Art

[0002] At present, water depth measurement mainly relies on model inversion techniques of optical remote sensing data, such as theoretical analysis, statistics, and semi-empirical and semi-theoretical models. Although these techniques can accurately obtain water depth information, they are limited by weather in emergency flood situations. In contrast, SAR technology can obtain ground data under various climatic conditions, so researchers are developing flood depth estimation methods based on SAR images. For example, Hostache et al. combined SAR images with high-resolution digital elevation models, estimated water levels using the principle of hydrodynamic coherence, and integrated the results into hydraulic models to reduce uncertainty. Mason et al. proposed a technique for automatically extracting subset water depths, by selecting water surface points in the image and refining these points using top-down clustering analysis, and using secondary reflections of water bodies and vegetation at the edge of the floodplain to calibrate water depth and location information. In order to monitor the depth of a large area of ​​floods, Matgen et al. used the height difference and critical success index of the nearest neighbor drainage network to convert the flood range map into a flood depth map, thereby achieving large-scale real-time monitoring. In China, with the frequent occurrence of flood disasters, risk management and hazard level assessment for different high-risk areas are particularly important. A sound flood risk management system can integrate disaster monitoring, prediction, and disaster prevention, mitigation, and relief measures to enhance the ability to respond to extreme disasters. Flood hazard maps and risk maps generated through multi-source data and statistical methods can reveal the possible impacts of floods of different levels. This enables emergency managers to formulate reasonable response strategies based on regional needs and disaster severity, optimize resource allocation, and thus reduce disaster losses.

[0003] Therefore, a flood depth extraction method based on remote sensing images is needed. Summary of the invention

[0004] The present invention provides a flood depth extraction method, system and electronic equipment based on remote sensing images to solve the problem of how to quickly and accurately determine the flood depth.

[0005] In order to solve the above problem, according to one aspect of the present invention, a flood depth extraction method based on remote sensing images is provided, the method comprising:

[0006] Water body extraction is performed based on the SAR image data of the target area and the trained water body extraction model to obtain water body extraction results;

[0007] Acquire original river data, and superimpose the original river data with the water body extraction result to obtain the flood area;

[0008] Acquire original DEM data of the target area, establish a hydrological relationship based on the original DEM data, and acquire a nearest neighbor drainage network relative height map of the target area;

[0009] The flood depth of the flooded area is determined based on the nearest neighbor drainage network relative height map, and a water depth map of the flooded area is generated.

[0010] Preferably, the method further comprises:

[0011] Preprocessing the SAR image data to obtain preprocessed SAR image data; wherein the preprocessing includes: image filtering, geocoding and radiation calibration;

[0012] Randomly select pre-processed SAR image data containing water bodies, set water body labels, and perform amplification to obtain training samples and test samples;

[0013] The TransUNet model is trained based on the training samples, and tested using the test samples to obtain a trained water body extraction model.

[0014] Preferably, the step of establishing a hydrological relationship based on the original DEM data and obtaining a nearest neighbor drainage network relative height map of the target area comprises:

[0015] Based on the original DEM data, depressions are identified and exit points are determined, and based on the exit points, depressions are filled in the original DEM data to obtain DEM data after depressions are filled;

[0016] Determine the flow direction of each pixel based on the DEM data after the depression is filled, and determine the LDD grid based on the flow direction of each pixel;

[0017] Establishing a flow direction-river network association network for classification based on the LDD grid and existing river network data;

[0018] Taking the flow direction-river network association network as a reference, the elevation difference between each pixel and the river network point in all similar pixels is calculated to obtain the relative height of each pixel point, and the nearest neighbor drainage network relative height map is obtained based on the relative height.

[0019] Preferably, the determining the flow direction of each pixel point based on the DEM data after the depression is filled includes:

[0020] For any pixel point, in the eight neighborhoods contained in the eight directions N, NE, E, SE, S, SW, W, and NW, the direction with the largest downward gradient with the central pixel is taken as the flow direction of the pixel; among which, the null value represents that the flow direction of the point is undefined. The neighborhood N(<i, j>) of the pixel point <i, j> is: N(<i, j>) = {<i-1, j-1>, <i-1, j>, <i-1, j+1>, <i, j+1>, <i+1, j-1>, <i+1, j>, <i+1, j+1>}, where i and j are the horizontal and vertical coordinates of the pixel point.

[0021] Preferably, the method calculates the flow direction of each pixel point in the following manner, including:

[0022] F(<i,j>)=<i-1,j-1>+Δ(LDD(<i,j>)),

[0023] Among them, F(<i,j>) represents the flow direction of pixel point <i,j>; Δ represents the relative position of pixel point <i,j> between neighboring pixels in a certain direction; i and j are the horizontal and vertical coordinates of the pixel point.

[0024] Preferably, the method obtains the relative height of each pixel point in the following manner, including:

[0025]

[0026] HEIGHT(<i,j>) is the relative height of pixel <i,j>; i and j are the horizontal and vertical coordinates of the pixel; H(<i,j>) is the elevation of the point at position <i,j> in the original DEM data; H D (<i,j>) represents the elevation of the grid point connected to point <i,j> by the flow direction.

[0027] Preferably, the step of determining the flood depth of the flooded area based on the nearest neighbor drainage network relative height map and generating a water depth map of the flooded area comprises:

[0028] Performing image segmentation on the nearest neighbor drainage network relative height map, calculating the values ​​of TP, TN, FP and FN according to the segmentation results, calculating MCC based on TP, TN, FP and FN, and taking the threshold corresponding to the maximum MCC as the optimal threshold;

[0029] For any pixel point in the flooded area, the optimal threshold is subtracted from the value of the pixel point to obtain the depth value of the pixel point, and a final water depth map of the flooded area is generated based on the depth value of each pixel point.

[0030] According to another aspect of the present invention, a flood depth extraction system based on remote sensing images is provided, the system comprising:

[0031] A water body extraction unit is used to extract water bodies based on SAR image data of the target area and a trained water body extraction model to obtain water body extraction results;

[0032] A flood area acquisition unit, used to acquire original river data, and superimpose the original river data and the water body extraction result to acquire the flood area;

[0033] A relative height map acquisition unit is used to acquire original DEM data of the target area, establish a hydrological relationship based on the original DEM data, and acquire a nearest neighbor drainage network relative height map of the target area;

[0034] The flood area water depth map generating unit is used to determine the flood depth of the flood area based on the nearest neighbor drainage network relative height map and generate the flood area water depth map.

[0035] Preferably, the system further comprises: a water body extraction model determination unit, configured to:

[0036] Preprocessing the SAR image data to obtain preprocessed SAR image data; wherein the preprocessing includes: image filtering, geocoding and radiation calibration;

[0037] Randomly select pre-processed SAR image data containing water bodies, set water body labels, and perform amplification to obtain training samples and test samples;

[0038] The TransUNet model is trained based on the training samples, and tested using the test samples to obtain a trained water body extraction model.

[0039] Preferably, the relative height map acquisition unit establishes a hydrological relationship based on the original DEM data to acquire a nearest neighbor drainage network relative height map of the target area, including:

[0040] Based on the original DEM data, depressions are identified and exit points are determined, and based on the exit points, depressions are filled in the original DEM data to obtain DEM data after depressions are filled;

[0041] Determine the flow direction of each pixel based on the DEM data after the depression is filled, and determine the LDD grid based on the flow direction of each pixel;

[0042] Establishing a flow direction-river network association network for classification based on the LDD grid and existing river network data;

[0043] Taking the flow direction-river network association network as a reference, the elevation difference between each pixel and the river network point in all similar pixels is calculated to obtain the relative height of each pixel point, and the nearest neighbor drainage network relative height map is obtained based on the relative height.

[0044] Preferably, the relative height map acquisition unit determines the flow direction of each pixel point based on the DEM data after the depression is filled, including:

[0045] For any pixel point, in the eight neighborhoods contained in the eight directions N, NE, E, SE, S, SW, W, and NW, the direction with the largest downward gradient with the central pixel is taken as the flow direction of the pixel; among which, the null value represents that the flow direction of the point is undefined. The neighborhood N(<i, j>) of the pixel point <i, j> is: N(<i, j>) = {<i-1, j-1>, <i-1, j>, <i-1, j+1>, <i, j+1>, <i+1, j-1>, <i+1, j>, <i+1, j+1>}, where i and j are the horizontal and vertical coordinates of the pixel point.

[0046] Preferably, the relative height map acquisition unit calculates the flow direction of each pixel point in the following manner, including:

[0047] F(<i,j>)=<i-1,j-1>+Δ(LDD(<i,j>)),

[0048] Among them, F(<i,j>) represents the flow direction of pixel point <i,j>; Δ represents the relative position of pixel point <i,j> between neighboring pixels in a certain direction; i and j are the horizontal and vertical coordinates of the pixel point.

[0049] Preferably, the relative height map acquisition unit acquires the relative height of each pixel point in the following manner, including:

[0050]

[0051] HEIGHT(<i,j>) is the relative height of pixel <i,j>; i and j are the horizontal and vertical coordinates of the pixel; H(<i,j>) is the elevation of the point at position <i,j> in the original DEM data; H D (<i,j>) represents the elevation of the grid point connected to point <i,j> by the flow direction.

[0052] Preferably, the flood area water depth map generating unit determines the flood depth of the flood area based on the nearest neighbor drainage network relative height map and generates the flood area water depth map, including:

[0053] Performing image segmentation on the nearest neighbor drainage network relative height map, calculating the values ​​of TP, TN, FP and FN according to the segmentation results, calculating MCC based on TP, TN, FP and FN, and taking the threshold corresponding to the maximum MCC as the optimal threshold;

[0054] For any pixel point in the flooded area, the optimal threshold is subtracted from the value of the pixel point to obtain the depth value of the pixel point, and a final water depth map of the flooded area is generated based on the depth value of each pixel point.

[0055] Based on another aspect of the present invention, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any step of a method for extracting flood depth based on remote sensing images.

[0056] According to another aspect of the present invention, the present invention provides an electronic device, including:

[0057] The computer-readable storage medium described above; and

[0058] One or more processors are used to execute the program in the computer-readable storage medium.

[0059] The present invention provides a flood depth extraction method, system and electronic equipment based on remote sensing images, including: extracting water based on SAR image data of a target area and a trained water extraction model to obtain a water extraction result; obtaining original river data, and superimposing the original river data and the water extraction result to obtain a flooded area; obtaining original DEM data of the target area, and establishing a hydrological relationship based on the original DEM data to obtain a nearest neighbor drainage network relative height map of the target area; determining the flood depth of the flooded area based on the nearest neighbor drainage network relative height map, and generating a water depth map of the flooded area. The method of the present invention can efficiently and accurately determine the flood depth and generate a water depth map of the flooded area, and has multiple advantages in terms of computational efficiency, extraction accuracy, scope of application, data dependence, and operational difficulty. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] A more complete understanding of exemplary embodiments of the present invention may be obtained by referring to the following drawings:

[0061] Figure 1 is a flow chart of a flood depth extraction method 100 based on remote sensing images according to an embodiment of the present invention;

[0062] Figure 2 It is an overall architecture diagram of a flood depth extraction method based on remote sensing images according to an embodiment of the present invention;

[0063] Figure 3 A diagram showing the results of water extraction in the floodplain of the experimental area according to an embodiment of the present invention;

[0064] Figure 4 A schematic diagram of DEM data of an experimental area according to an embodiment of the present invention;

[0065] Figure 5 A relative height map of a floodplain according to an embodiment of the present invention;

[0066] Figure 6 A water depth map of a floodplain according to an embodiment of the present invention;

[0067] Figure 7 Schematic diagram of the structure of a flood depth extraction system 700 based on remote sensing images according to an embodiment of the present invention. DETAILED DESCRIPTION

[0068] Now, exemplary embodiments of the present invention are described with reference to the accompanying drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. These embodiments are provided to disclose the present invention in detail and completely and to fully convey the scope of the present invention to those skilled in the art. The terms used in the exemplary embodiments shown in the accompanying drawings are not intended to limit the present invention. In the accompanying drawings, the same units / elements are marked with the same reference numerals.

[0069] Unless otherwise specified, the terms (including technical terms) used herein have the commonly understood meanings to those skilled in the art. In addition, it is understood that the terms defined in commonly used dictionaries should be understood to have the same meanings as those in the context of the relevant fields, and should not be understood as idealized or overly formal meanings.

[0070] Although the current flood depth extraction algorithm based on SAR images can accurately extract flood inundation information, the algorithm's dependence on ground data, the complexity of the principle, and the computational efficiency make it unsuitable for large-scale flood monitoring. In response to the current demand for large-scale flood depth information extraction, the present invention proposes a rapid flood depth extraction method based on remote sensing images on the basis of a rapid flood range extraction method. The method of the present invention first normalizes the terrain to establish a hydrological association between different areas in the image and adjacent rivers and lakes. Then, the original river data is overlaid with the water body results extracted based on the TransUNet model to obtain the flooded area, and then calculations are performed based on the flooded area and the hydrological association results to further obtain the water depth results of the flooded area.

[0071] Figure 1 FIG. 1 is a flow chart of a flood depth extraction method 100 based on remote sensing images according to an embodiment of the present invention. Figure 1 As shown, the flood depth extraction method based on remote sensing images provided by the embodiment of the present invention can efficiently and accurately determine the flood depth and generate a water depth map of the flooded area, and has multiple advantages in terms of computational efficiency, extraction accuracy, scope of application, data dependence, and operational difficulty. The flood depth extraction method 100 based on remote sensing images provided by the embodiment of the present invention starts from step 101, and performs water extraction based on the SAR image data of the target area and the trained water extraction model to obtain a water extraction result.

[0072] Preferably, the method further comprises:

[0073] Preprocessing the SAR image data to obtain preprocessed SAR image data; wherein the preprocessing includes: image filtering, geocoding and radiation calibration;

[0074] Randomly select pre-processed SAR image data containing water bodies, set water body labels, and perform amplification to obtain training samples and test samples;

[0075] The TransUNet model is trained based on the training samples, and tested using the test samples to obtain a trained water body extraction model.

[0076] In step 102, original river data is obtained, and the original river data and the water body extraction result are superimposed to obtain the flood area.

[0077] Combination Figure 2 As shown, in the present invention, after acquiring the SAR image data of the target area, the SAR impact data is input into the trained water body extraction model for water body extraction to obtain the water body extraction result, and the water body extraction result and the original river data are superimposed to obtain the flooded area.

[0078] Among them, in the field of flood mapping, water bodies are usually divided into permanent water bodies and transient water bodies. Floods belong to transient water bodies, so the flood range extraction results do not include the original river channels, lakes and other areas. In order to better perform overlay analysis with the relative height map, the present invention combines the original river channel data with the extracted flood range map to obtain the flood area extraction results including permanent water bodies and transient water bodies.

[0079] In the present invention, the process of obtaining a water body extraction model includes:

[0080] 1) Image data preprocessing, including image filtering, geocoding, and radiation calibration of SAR images.

[0081] 2) Preparation of SAR image water body sample dataset. After the SAR image data is preprocessed, SAR image samples containing water bodies are randomly selected, water body labels are set, and the samples are amplified. Finally, the samples are divided into training set and test set.

[0082] 3) Perform model training and verification, use the TransUNet model to extract water bodies, and obtain a trained water body extraction model. The TransUNet structure is designed based on the classic U-Net, which is characterized by a U-shaped encoder and decoder combination. In TransUNet, the encoder first uses a convolution method to extract shallow features of the image. These shallow features are linearly embedded and positionally encoded to obtain labeled image blocks. Subsequently, these labeled image blocks are transmitted to multiple Transformer modules, thereby capturing local spatial details in the convolution stage and capturing global context information in the Transformer stage.

[0083] In step 103, the original DEM data of the target area is obtained, and a hydrological relationship is established based on the original DEM data to obtain a nearest neighbor drainage network relative height map of the target area.

[0084] Preferably, the step of establishing a hydrological relationship based on the original DEM data and obtaining a nearest neighbor drainage network relative height map of the target area comprises:

[0085] Based on the original DEM data, depressions are identified and exit points are determined, and based on the exit points, depressions are filled in the original DEM data to obtain DEM data after depressions are filled;

[0086] Determine the flow direction of each pixel based on the DEM data after the depression is filled, and determine the LDD grid based on the flow direction of each pixel;

[0087] Establishing a flow direction-river network association network for classification based on the LDD grid and existing river network data;

[0088] Taking the flow direction-river network association network as a reference, the elevation difference between each pixel and the river network point in all similar pixels is calculated to obtain the relative height of each pixel point, and the nearest neighbor drainage network relative height map is obtained based on the relative height.

[0089] Preferably, the determining the flow direction of each pixel point based on the DEM data after the depression is filled includes:

[0090] For any pixel point, in the eight neighborhoods contained in the eight directions N, NE, E, SE, S, SW, W, and NW, the direction with the largest downward gradient with the central pixel is taken as the flow direction of the pixel; among which, the null value represents that the flow direction of the point is undefined. The neighborhood N(<i, j>) of the pixel point <i, j> is: N(<i, j>) = {<i-1, j-1>, <i-1, j>, <i-1, j+1>, <i, j+1>, <i+1, j-1>, <i+1, j>, <i+1, j+1>}, where i and j are the horizontal and vertical coordinates of the pixel point.

[0091] Preferably, the method calculates the flow direction of each pixel point in the following manner, including:

[0092] F(<i,j>)=<i-1,j-1>+Δ(LDD(<i,j>)),

[0093] Among them, F(<i,j>) represents the flow direction of pixel point <i,j>; Δ represents the relative position of pixel point <i,j> between neighboring pixels in a certain direction; i and j are the horizontal and vertical coordinates of the pixel point.

[0094] Preferably, the method obtains the relative height of each pixel point in the following manner, including:

[0095]

[0096] HEIGHT(<i,j>) is the relative height of pixel <i,j>; i and j are the horizontal and vertical coordinates of the pixel; H(<i,j>) is the elevation of the point at position <i,j> in the original DEM data; H D (<i,j>) represents the elevation of the grid point connected to point <i,j> by the flow direction.

[0097] In the present invention, it is also necessary to establish a hydrological relationship based on the original DEM data and obtain a nearest neighbor drainage network relative height map of the target area. Specifically, the process of determining the relative height includes:

[0098] 1) Depression filling: In order to avoid the interference of abnormal areas such as noise and forest tree gaps in the calculation of flow direction as much as possible, the original DEM data must first be "depression filled" to raise the internal height of the "depression" to the elevation of the "exit point" (the lowest point on the edge of the depression).

[0099] 2) Calculate the flow direction. Use the filled DEM data to determine the flow direction of each pixel. In the eight neighborhoods contained in the eight directions of N, NE, E, SE, S, SW, W, and NW, the direction with the largest downward gradient from the center pixel is the flow direction of the pixel. Among them, the null value means that the flow direction of the point is undefined. The neighborhood N(<i, j>) of the pixel <i, j> is:

[0100] N(<i, j>) = {<i - 1, j - 1>, <i - 1, j>, <i - 1, j + 1>, <i, j + 1>, <i + 1, j - 1>, <i + 1, j>, <i + 1, j + 1>}

[0101] Among them, calculating the flow direction of each pixel point includes:

[0102] F(<i, j>) = <i - 1, j - 1> + Δ(LDD(<i, j>))

[0103] Among them, F(<i, j>) represents the flow direction of the pixel point <i, j>; Δ represents the relative position between the pixel point <i, j> and the neighboring pixels in a certain direction; i and j are the horizontal and vertical coordinates of the pixel point.

[0104] 3) Calculate the LDD grid. By traversing the DEM image, the flow direction data of the study area is obtained, that is, the LDD grid.

[0105] 4) Flow direction - river network association network. According to the existing river network data and the LDD grid, a flow direction - river network association network for classification can be established. Each pixel is connected to the nearest river network point along the "flow direction", so as to realize the classification of the pixels around the river network.

[0106] 5) Calculate the elevation difference. Taking the flow direction - river network association network as a reference, calculate the elevation difference between each pixel and the river network point among all pixels of the same type, so as to obtain the relative height map of the nearest neighbor drainage network.

[0107] Among them, the relative height of each pixel point is obtained by the following method, including:

[0108]

[0109] Among them, H(<i, j>) is the elevation of the point located at <i, j> in the original DEM image, and H D (<i, j>) represents the elevation of the grid point connected to the point <i, j> through the flow direction.

[0110] In step 104, based on the relative height map of the nearest neighbor drainage network, determine the flood depth of the flood area and generate a water depth map of the flood area.

[0111] Preferably, the determining the flood depth of the flood area based on the relative height map of the nearest neighbor drainage network and generating a water depth map of the flood area includes:

[0112] Performing image segmentation on the nearest neighbor drainage network relative height map, calculating the values ​​of TP, TN, FP and FN according to the segmentation results, calculating MCC based on TP, TN, FP and FN, and taking the threshold corresponding to the maximum MCC as the optimal threshold;

[0113] For any pixel point in the flooded area, the optimal threshold is subtracted from the value of the pixel point to obtain the depth value of the pixel point, and a final water depth map of the flooded area is generated based on the depth value of each pixel point.

[0114] In the present invention, the nearest neighbor drainage network relative height Figure 2 Classification. After obtaining the relative height map of the study area, the relative height image is segmented, and the area less than the threshold is extracted as the flood area prediction result. Often, the larger the threshold, the larger the predicted flood range and the deeper the flooding depth. In the present invention, when MCC is the largest, the HEIGHT value corresponding to MCC is the optimal threshold HEIGHT T .

[0115] MCC is a statistic used to evaluate the performance of a binary classification model. Segmentation of relative height images into flooded areas (positive class) and non-flooded areas (negative class) is a binary classification problem. The value range of MCC is from -1 to 1, where 1 indicates a completely correct classification, 0 indicates that the classification effect is no better than random guessing, and -1 indicates a completely wrong classification. It comprehensively considers the four indicators of true positive (True Positive, TP), true negative (True Negative, TN), false positive (False Positive, FP) and false negative (False Negative, FN). The calculation formula is:

[0116]

[0117] Among them, TP is true positive, FP is false positive, TN is true negative, and FN is false negative.

[0118] When evaluating floodplain prediction results, true positives are areas that are actually floodplains and are correctly predicted to be floodplains; true negatives are areas that are not actually floodplains and are correctly predicted to be nonfloodplains; false positives are areas that are not actually floodplains but are predicted to be floodplains; and false negatives are areas that are actually floodplains but are predicted to be nonfloodplains.

[0119] After obtaining the optimal segmentation threshold, use the formula Depth(<i,j>)=HEIGHT T -HEIGHT(<i,j>), calculate the depth value of the pixel, HEIGHT Trepresents the optimal threshold, that is, the optimal threshold is subtracted from the numerical value of all pixels in the predicted flood map to obtain the depth value of the pixel; then, the pixels in the flood area are extracted to generate the final flood depth map.

[0120] In the present invention, 7 SAR images of COSMO-SkyMed satellite in 2019-07, 2019-09, 2019-11, 2019-12, 2020-02, 2020-05, and 2020-06 are used for calculation. The basic parameters of the images are shown in Table 1. The present invention uses SAR images in July 2019 to make training samples, and uses SAR images in May 2020 to verify the accuracy of water body extraction, wherein the geocoded DEM data uses TanDEM-X90m data. Finally, the water body extraction results in June 2020 are used for water body depth calculation. During the calculation process, the water body extraction results of the floodplain in the experimental area are obtained as shown in the figure. Figure 3 As shown in the figure, the DEM data of the experimental area used is as follows Figure 4 The relative height map of the floodplain is shown in Figure 5 The final floodplain water depth map is shown in Figure 6 shown.

[0121] Table 1 Image parameter table

[0122] satellite Cosmo-SkyMed polarization HH Resolution / m 3 track Raise track Angle of incidence / (°) 26.6 Width / Km 40×40

[0123] Figure 7 FIG. 7 is a schematic diagram of a flood depth extraction system 700 based on remote sensing images according to an embodiment of the present invention. Figure 7 As shown, a flood depth extraction system 700 based on remote sensing images provided by an embodiment of the present invention includes: a water body extraction unit 701, a flooded area acquisition unit 702, a relative height map acquisition unit 703 and a flooded area water body depth map generation unit 704.

[0124] Preferably, the water body extraction unit 701 is used to perform water body extraction based on the SAR image data of the target area and a trained water body extraction model to obtain a water body extraction result.

[0125] Preferably, the system further comprises: a water body extraction model determination unit, configured to:

[0126] Preprocessing the SAR image data to obtain preprocessed SAR image data; wherein the preprocessing includes: image filtering, geocoding and radiation calibration;

[0127] Randomly select pre-processed SAR image data containing water bodies, set water body labels, and perform amplification to obtain training samples and test samples;

[0128] The TransUNet model is trained based on the training samples, and tested using the test samples to obtain a trained water body extraction model.

[0129] Preferably, the flood area acquisition unit 702 is used to acquire original river data, and superimpose the original river data and the water body extraction result to acquire the flood area.

[0130] Preferably, the relative height map acquisition unit 703 is used to acquire original DEM data of the target area, establish a hydrological relationship based on the original DEM data, and acquire a nearest neighbor drainage network relative height map of the target area.

[0131] Preferably, the relative height map acquisition unit 703 establishes a hydrological relationship based on the original DEM data to acquire a nearest neighbor drainage network relative height map of the target area, including:

[0132] Based on the original DEM data, depressions are identified and exit points are determined, and based on the exit points, depressions are filled in the original DEM data to obtain DEM data after depressions are filled;

[0133] Determine the flow direction of each pixel based on the DEM data after the depression is filled, and determine the LDD grid based on the flow direction of each pixel;

[0134] Establishing a flow direction-river network association network for classification based on the LDD grid and existing river network data;

[0135] Taking the flow direction-river network association network as a reference, the elevation difference between each pixel and the river network point in all similar pixels is calculated to obtain the relative height of each pixel point, and the nearest neighbor drainage network relative height map is obtained based on the relative height.

[0136] Preferably, the relative height map acquisition unit 703 determines the flow direction of each pixel point based on the DEM data after the depression is filled, including:

[0137] For any pixel point, in the eight neighborhoods contained in the eight directions N, NE, E, SE, S, SW, W, and NW, the direction with the largest downward gradient with the central pixel is taken as the flow direction of the pixel; the null value represents that the flow direction of the point is undefined, and the neighborhood N(<i, j>) of the pixel point <i, j> is: N(<i, j>) = {<i-1, j-1>, <i-1, j>, <i-1, j+1>, <i, j+1>, <i+1, j-1>, <i+1, j>, <i+1 , j+1>}, i and j are the horizontal and vertical coordinates of the pixel point.

[0138] Preferably, the relative height map acquisition unit 703 calculates the flow direction of each pixel point in the following manner, including:

[0139] F(<i,j>) = <i - 1,j - 1> + Δ(LDD(<i,j>)),

[0140] where F(<i,j>) represents the flow direction of the pixel point <i,j>; Δ represents the relative position between the pixel point <i,j> and the neighboring pixels in a certain direction; i and j are the horizontal and vertical coordinates of the pixel point.

[0141] Preferably, the relative height map acquisition unit 703 obtains the relative height of each pixel point in the following manner, including:

[0142]

[0143] where HEIGHT(<i,j>) is the relative height of the pixel point <i,j>; i and j are the horizontal and vertical coordinates of the pixel point; H(<i,j>) is the elevation of the point at the position <i,j> in the original DEM data; H D (<i,j>) represents the elevation of the grid point connected to the point <i,j> through the flow direction.

[0144] Preferably, the flood area water depth map generation unit 704 is used to determine the flood depth of the flood area based on the relative height map of the nearest neighbor drainage network and generate a flood area water depth map.

[0145] Preferably, the flood area water depth map generation unit 704 determines the flood depth of the flood area based on the relative height map of the nearest neighbor drainage network and generates a flood area water depth map, including:

[0146] Performing image segmentation on the relative height map of the nearest neighbor drainage network, calculating the values of TP, TN, FP, and FN according to the segmentation result, calculating MCC based on TP, TN, FP, and FN, and using the threshold corresponding to the maximum MCC as the optimal threshold;

[0147] For any pixel point in the flood area, subtracting the value of the pixel point from the optimal threshold to obtain the depth value of the pixel point, and generating the final flood area water depth map based on the depth values of each pixel point.

[0148] The flood depth extraction system 700 based on remote sensing images in the embodiments of the present invention corresponds to the flood depth extraction method 100 based on remote sensing images in another embodiment of the present invention, which will not be elaborated here.

[0149] Based on another aspect of the present invention, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any step of a method for extracting flood depth based on remote sensing images.

[0150] According to another aspect of the present invention, the present invention provides an electronic device, including:

[0151] The computer-readable storage medium described above; and

[0152] One or more processors are used to execute the program in the computer-readable storage medium.

[0153] The invention has been described above with reference to a few embodiments. However, it is readily apparent to a person skilled in the art that other embodiments than the ones disclosed above are equally within the scope of the invention, as defined by the appended patent claims.

[0154] Generally, all terms used in the claims are to be interpreted according to their ordinary meaning in the technical field, unless explicitly defined otherwise therein. All references to "a / said / the [means, components, etc.]" are to be openly interpreted as at least one instance of said means, components, etc., unless explicitly stated otherwise. The steps of any method disclosed herein do not necessarily have to be performed in the exact order disclosed, unless explicitly stated otherwise.

[0155] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0156] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0157] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0158] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0159] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A flood depth extraction method based on remote sensing images, characterized in that: The method includes: Performing water body extraction based on the SAR image data of the target area and the trained water body extraction model to obtain a water body extraction result; Obtaining the original river channel data, and superimposing the original river channel data and the water body extraction result to obtain a flood area; Obtaining the original DEM data of the target area, establishing a hydrological relationship based on the original DEM data, and obtaining the relative height map of the nearest neighbor drainage network in the target area; Determining the flood depth of the flood area based on the relative height map of the nearest neighbor drainage network and generating a water depth map of the flood area.

2. The method according to claim 1, characterized in that The method further includes: Preprocessing the SAR image data to obtain preprocessed SAR image data; wherein, the preprocessing includes: image filtering, geocoding, and radiometric calibration; Randomly selecting preprocessed SAR image data containing water bodies to set water body labels and performing amplification to obtain training samples and test samples; Training the TransUNet model based on the training samples and testing with the test samples to obtain a trained water body extraction model.

3. The method according to claim 1, characterized in that The establishing a hydrological relationship based on the original DEM data and obtaining the relative height map of the nearest neighbor drainage network in the target area includes: Identifying depressions based on the original DEM data, determining outlet points, and filling the depressions in the original DEM data based on the outlet points to obtain a DEM data after depression filling; Determining the flow direction of each pixel point based on the DEM data after depression filling, and determining the LDD grid based on the flow direction of each pixel point; Establishing a flow direction - river network association network for classification based on the LDD grid and the existing river network data; Taking the flow direction - river network association network as a reference, calculating the elevation difference between each pixel and the river network points among all pixels of the same type, obtaining the relative height of each pixel point, and obtaining the relative height map of the nearest neighbor drainage network based on the relative height.

4. The method according to claim 3, characterized in that The determining the flow direction of each pixel point based on the DEM data after depression filling includes: For any pixel point, among the eight neighborhoods included in the eight directions of N, NE, E, SE, S, SW, W, NW, the direction with the largest downward gradient between the center pixel and the neighboring pixels is taken as the flow direction of the pixel; where null values represent that the flow direction of this point is undefined, and the neighborhood N(<i,j>) of the pixel point <i,j> is: N(<i,j>) = {<i - 1,j - 1>,<i - 1,j>,<i - 1,j + 1>,<i,j + 1>,<i + 1,j - 1>,<i + 1,j>,<i + 1,j + 1>}, and i and j are the horizontal and vertical coordinates of the pixel point.

5. The method according to claim 3, characterized in that: The method calculates the flow direction of each pixel point in the following manner, including: F(<i,j>) = <i - 1,j - 1> + Δ(LDD(<i,j>)), where F(<i,j>) represents the flow direction of the pixel point <i,j>; Δ represents the relative position between the pixel point <i,j> and the neighboring pixels in a certain direction; and i and j are the horizontal and vertical coordinates of the pixel point.

6. The method according to claim 3, characterized in that The method obtains the relative height of each pixel point in the following manner, including: Among them, HEIGHT(<i,j> ) is a pixel<i,j> The relative height of the pixel; i and j are the horizontal and vertical coordinates of the pixel; H(<i,j> ) is the original DEM data located at<i,j> The elevation of the point at which the position is located; H D (<i,j> ) indicates the flow direction and point<i,j> The connected grid point elevations.

7. The method according to claim 1, characterized in that The method of determining the flood depth of the flooded area based on the nearest neighbor drainage network relative height map and generating a water depth map of the flooded area includes: Performing image segmentation on the nearest neighbor drainage network relative height map, calculating the values ​​of TP, TN, FP and FN according to the segmentation results, calculating MCC based on TP, TN, FP and FN, and taking the threshold corresponding to the maximum MCC as the optimal threshold; For any pixel point in the flooded area, the optimal threshold is subtracted from the value of the pixel point to obtain the depth value of the pixel point, and a final water depth map of the flooded area is generated based on the depth value of each pixel point.

8. A flood depth extraction system based on remote sensing images, characterized in that: The system comprises: A water body extraction unit is used to extract water bodies based on SAR image data of the target area and a trained water body extraction model to obtain water body extraction results; A flood area acquisition unit, used to acquire original river data, and superimpose the original river data and the water body extraction result to acquire the flood area; A relative height map acquisition unit is used to acquire original DEM data of the target area, establish a hydrological relationship based on the original DEM data, and acquire a nearest neighbor drainage network relative height map of the target area; The flood area water depth map generating unit is used to determine the flood depth of the flood area based on the nearest neighbor drainage network relative height map and generate the flood area water depth map.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. An electronic device, characterized in that: include: The computer readable storage medium as claimed in claim 9; as well as One or more processors are used to execute the program in the computer-readable storage medium.

Images